AI memory

Let AI steps in your automations remember context from earlier runs.

Today, each Prompt only sees its current input. It doesn’t keep a customer thread, earlier notes, or the result of the last run, so every run starts from zero. We’re working on Memory to change that.

Memory will let AI components keep a short history of the questions they receive and the answers they give. When you turn it on for a step, that history is passed into later runs, so the AI can answer with the earlier context in mind. You’ll be able to decide whether the context is kept for a single automation or shared across your workspace.

This will help with things like support replies that follow a whole conversation, recurring reports that build on the previous one, and routing decisions that learn from how similar cases were handled before. Memory will be opt-in, so existing automations keep working exactly as they do today.